Ilnar Nizametdinov

Technology was always the pull. In late 2022 I left enterprise procurement to build software full-time.

Main project: PowerCV (v3.3.1) — an AI resume optimisation platform I rearchitected from an open-source fork into a real production system.

What I built:

Backend architecture

Async FastAPI application with Pydantic v2 validation, custom middleware, and global error handling. CVWorkflowOrchestrator — a custom async pipeline that takes a CV from intake through analysis, gap detection, content optimisation, ATS scoring, and cover letter generation. All steps tracked and observable.

Data layer

Migrated the database from MongoDB to PostgreSQL (asyncpg) — designed connection pooling, wrote migration scripts, and restructured every query. Added Redis caching to reduce LLM API costs. Both databases run alongside each other in production.

AI / LLM integration

LiteLLM abstraction layer over Cerebras, OpenAI, and Ollama — switch providers by changing one environment variable. LangChain for prompt chaining and retrieval patterns. Token usage is tracked per request.

Infrastructure and DevOps

Full GitHub Actions CI/CD pipeline: Ruff linting → Pytest with real service containers (PostgreSQL, Redis) → Docker multi-stage build. Kubernetes-ready deployment with /health probes, 12-factor config, and secrets via env vars. Sentry error tracking, structured JSON logging. n8n integration layer for external workflow automation. Playwright PDF generation via Typst templates. Job description scraper (BeautifulSoup, lxml, async HTTP) covering LinkedIn and Indeed.

Stack: Python 3.11 · FastAPI · PostgreSQL (asyncpg) · Redis · Kafka · Docker · Kubernetes · GitHub Actions · Sentry · LiteLLM · LangChain · Playwright · n8n · Pytest · Ruff · Pydantic v2

github.com/ILLnar-Nizami/PowerCV (branch: dev · v3.3.1 · 305+ commits)